Responsible AI is a tech driver of sustainable economic growth that protects democratic liberties. The systematic design, implementation, and deployment of AI for good are demanding tasks, given the diversity of those impacted. Engaging a representative sample of AI’s heterogeneous user base to gauge the benefits it expects requires innovative participatory activities interspersed throughout the stages of the AI development process. Translating stakeholder input from the jargon-free vocabulary in which it is collected to coherent, comprehensive, industry-standard artefacts that experts can use to build responsible AI in practice is also challenging. Developing a robust assessment framework with objective metrics for evaluating intelligent tech adds to the overall difficulty. We capture these aspects in seven key challenges which we address by proposing a novel, systematic, participatory approach to co-designing and co-assessing responsible AI. We apply the approach to architect an AI recommender system that supports transport authorities, industry, policymakers, and the public with their urban mobility decisions. Throughout three workshops, representatives from the four stakeholder categories worked with domain experts to co-develop a system blueprint featuring complementary tech from across the AI spectrum and an evaluation framework with robust blueprint assessment metrics. This paper presents the two artefacts alongside a detailed account of the innovative workshop activities leading to their co-creation.